Configuring Storage Tiers

To assign disk space to a storage tier, do this:

1. Assign all chunk servers with SSD drives to tier 1. You can do this when setting up a chunk server; see Creating the Chunk Server for details.

   Note: For information on recommended SSD drives, see Using SSD Drives .

2. Assign the frequently accessed directories and files to tier 1 with the pstorage set-attr command. For example:

   # pstorage set-attr -R /pstorage/pcs1/private/101 tier=1

   This command assigns the directory /pstorage/pcs1/private/101 and its contents to tier 1.

When assigning storage to tiers, have in mind that a tier number itself does not have any effect on drive performance. However, the general recommendation is to assign faster storage drives to lower tiers (see table below).

This recommendation is related to how Parallels Cloud Storage works with storage space. If a storage tier runs out of free space, Parallels Cloud Storage will attempt to temporarily use a lower tier, and if those are full too, a higher tier. If you add more storage to the original tier later, the data, temporarily stored elsewhere, will be moved to the original tier where it should have been stored originally.

For example, if you try to write data to the tier 2 and it is full, Parallels Cloud Storage will attempt to write that data to tier 1, then to tier 0, and then to tier 3. If you add more storage to tier 2 later, the aforementioned data, now stored on the tier 1, or 0, or 3, will be moved back to the tier 2 where it was meant to be stored originally.

Automatic Data Balancing

To maximize the I/O performance of storage drives working in a cluster, Parallels Cloud Storage automatically balances drive load by moving data chunks from drives under high load to those under less load as well as moving the most used ("hot") data chunks to faster drives and least used ("cold") data chunks to slower drives. This approach enables optimal utilization of drives of various capacity and performance working on the same tier in the same cluster.

For example, in a cluster that includes a 4 TB, a 1 TB, and a 500 GB drive, the 4 TB drive will normally have to handle the most I/O requests as it can store the most data chunks, while the 500 GB drive will normally have much fewer I/O requests to process as it can store the least data chunks. As a result, the 4 TB drive may become overloaded while the 500 GB drive will remain relatively idle. To avoid such performance imbalance, Parallels Cloud Storage will automatically try to offload the drive under the highest load (the 4 TB one in our example) by moving some of the data chunks from it to the drive under the least load (the 500 GB one in our example). The resulting data allocation pattern may look like this:

• 500 GB drive: ~6K chunks, 20% disk space free
• 1 TB drive: ~10K chunks, 30% disk space free
• 4 TB drive: ~12K chunks, 70% disk space free

In addition, Parallels Cloud Storage will automatically determine which data chunks are requested most and least and place the "hot" data chunks on the fastest drives (e.g., SAS, in a mixed cluster) and the "cold" data chunks on the slowest drives (e.g., SATA, in a mixed cluster).